Credit Decisions, Affordability Models and Fair Treatment

When financial institutions use artificial intelligence in credit decisions, they must ensure these systems operate fairly and in line with existing UK regulatory expectations. The FCA expects firms to apply appropriate governance frameworks that support responsible AI usage. This includes ensuring credit decisions are made through systems that are not only accurate but also fair, transparent, and aligned with consumer protection principles. The AI must not result in discriminatory outcomes or unreasonably disadvantage certain groups of applicants.

Affordability Models and AI

Affordability models play a central role in credit decisions. These models assess whether a borrower can reasonably meet their repayment obligations. AI systems used in these models must reflect the borrower’s true financial position, including income, expenses, and existing debts. Firms must ensure these models do not rely on data that could lead to unfair treatment. For example, an AI system that makes assumptions about income based on postcode or demographic data could unintentionally discriminate against certain groups. The FCA expects firms to validate these models through testing and monitoring to ensure they align with affordability principles.

• Ensure models are tested for bias across demographic groups
• Regularly review data sources to prevent discriminatory outcomes
• Document how affordability is calculated and reviewed

Fair Treatment and AI Decision-Making

AI systems must not produce outcomes that are unfair or discriminatory. The EU AI Act, which applies in the UK through domestic legislation, prohibits certain AI practices that could lead to unfair treatment. These include systems that make decisions based on sensitive characteristics such as race, gender, or religion. Firms must ensure that credit decisions are made through systems that are explainable, auditable, and aligned with fair treatment principles. Where AI is used, firms must be able to explain how decisions are reached, particularly when they affect consumers negatively.

For example, a credit scoring model that uses data from social media or other non-traditional sources must be carefully reviewed to ensure it does not unintentionally penalise individuals from certain backgrounds. The FCA expects firms to have processes in place to monitor these models for fairness and take corrective action if disparities are identified. This includes conducting regular audits of AI systems and maintaining records of these checks.

• Implement fairness monitoring tools to detect bias
• Ensure credit decisions are explainable to consumers
• Train staff on fair treatment principles and AI use

As of 2 February 2025, firms must comply with the EU AI Act’s Article 5, which prohibits certain AI practices that could lead to unfair treatment. The FCA expects firms to apply these principles through their AI governance frameworks. The Digital Omnibus on AI, which came into force on 27 July 2026, further clarifies these obligations. Firms must ensure that AI systems used in credit decisions meet these standards, including through machine-readable marking for generative AI systems placed on the market before 2 August 2026.

ISO/IEC 42001:2023 provides a framework for AI management systems. Firms should consider aligning their AI governance with this standard to ensure they meet international best practices. The standard requires firms to have processes for AI risk management, including through the establishment of AI governance structures. The first UKAS-accredited certification body for AI management systems was BSI, which became certified on 15 January 2026. Firms may choose to pursue certification to demonstrate their commitment to responsible AI use.

By embedding these principles into credit decision-making processes, firms can reduce the risk of regulatory scrutiny and ensure that AI systems support fair and responsible lending. The focus must always be on protecting consumers while enabling effective credit assessment through AI. Regular training, testing, and monitoring of AI systems are essential to maintain compliance and uphold consumer trust.